User throughput, which is one of the most important performance indicator in the mobile network is a function of SINR, available PRBs (Physical Resource Blocks) and spectrum efficiency (SE).
In the threshold based optimization feature of MOON, both SINR and available resource, PRBs are the target performance to improve. This feature identifies low user throughput cells based on network KPI data as well as field test data when it is available. Then, this feature automatically identifies the root causes of low throughput and the problematic cells or target cells to optimize are categorized into 5 groups, overshooting cell, undershooting cell, non-dominant cell, traffic unbalanced cell and high traffic cell. Corresponding to the categories, different but suitable solutions such as RF tuning, interFreq. or intraFreq. offloading, handover, CA (Carrier Aggregation) configuration and UL quality related system parameters are developed and proposed.
Our unique approach especially regarding RF tuning is to utilize G-factor as an indicator of robustness toward quality mobile network. G-Factor identifies discrete overlapping areas and provides target area for focused RF optimization. G-factor doesn’t depend upon the traffic load of the own & neighbour cells, hence is same throughout the 24 hr period. Higher G-factor means the LTE network is more robust against traffic load. MOON identifies low G-factor spots automatically and proposes antenna tilt, azimuth and RS offset changes using Genetic algorithm (GA). GA is one of the most useful meta-heuristic algorithms to solve various optimization problems where finding local or global minimum or maximum is very difficult or time consuming. This feature will save lots of manual RF tuning time and shorten the network performance improvement timeline.
The suitable offloading target cells are identified based on relationship with source cell in terms of geographical distance and handover attempts and PRB loads of the target cells. MOON presents the candidate target cells for offloading and provides the simulation result showing new PRB load situation at the source cell and the target cells or target RF layers in case interFreq. offloading is selected.
VoLTE Audio Gap (Mute) Detection and Analysis
VoLTE mute (audio gap) is a critical end user experience problem but difficult to detect from network KPI or even from drive test. In the conventional VoLTE mute field test, the problem analysis and counter-actions are developed only for detected VoLTE mutes. This will require exhaustive field tests because new VoLTE mutes may occur whenever new field test is performed.
As an innovative approach, MOON is equipped with deep learning (Recurrent Neural Network) algorithm to predict VoLTE mutes. The uniqueness of this approach is as follow.
- Simplify drive test for VoLTE mute detection as it requires only RF scanner or UE’s ping test after learning process.
- The deep learning algorithm predicts VoLTE mutes and provides the root causes for proper improvement of VoLTE quality.